An arch dam discharge vibration monitoring system and a method of identifying damage to a discharge structure

Abstract

The application provides an arch dam discharge vibration monitoring system and a method for identifying discharge structure damage, comprising a vibration sensor, a power module, a communication module, an arch dam discharge vibration measuring point arrangement design module, a vibration monitoring module and an upper computer; the vibration sensor is used for monitoring a discharge vibration signal in real time to obtain a vibration monitoring signal; the power module is used for power supply for other modules; the communication module is used for real-time transmission of the vibration monitoring signal; the arch dam discharge vibration measuring point arrangement design module is used for determining the installation position of the vibration sensor on site; the vibration monitoring module and the upper computer are used for data calculation, analysis and storage of the vibration monitoring signal; through self-learning of the flood discharge vibration data, the normal threshold of the vibration index is continuously updated and optimized, the prediction and early warning accuracy of abnormal conditions of the flood discharge structure is improved; the vibration monitoring result analysis report is automatically generated, and whether the discharge structure building operation is normal or not is accurately analyzed and judged.

Description

Technical Field

[0001] This invention relates to the field of arch dam monitoring technology, and more specifically, to an arch dam spillway vibration monitoring system and a method for identifying spillway structural damage. Background Technology

[0002] With the rapid development and maturation of infrastructure technology, dam construction has progressed rapidly and is becoming increasingly taller, with ultra-high arch dams of 200-300 meters appearing one after another. These ultra-high arch dams are characterized by large flood flows, complex dam layouts, and high head pressure. They are mostly equipped with spillway gates for flood discharge, and the discharge of floodwater involves the transfer and release of energy, inevitably impacting the flow structures and energy dissipation facilities. Therefore, flood discharge safety monitoring and damage identification of the flow structures and lower energy dissipation facilities of ultra-high arch dams are particularly important. Current flood discharge vibration monitoring technologies mostly use voltage-type vibration sensors or pressure sensors, which suffer from poor anti-interference capabilities, poor insulation, severe signal distortion, low accuracy, short service life, and susceptibility to damage in humid discharge environments, leading to monitoring system failure. Furthermore, existing methods for monitoring damage to discharge structures often result in erroneous judgments due to inaccurate threshold settings, leading to false alarms under normal conditions and failure to alarm under abnormal conditions.

[0003] In view of this, the present invention proposes an arch dam spillway vibration monitoring system and a method for identifying spillway structure damage. By using spillway vibration data for self-learning, the system continuously updates and optimizes the normal threshold of vibration indicators. As the amount of spillway vibration monitoring data increases, the vibration indicator threshold will continuously approach the true threshold, greatly improving the accuracy of forecasting and early warning of abnormal situations in the spillway structure. It can also automatically generate vibration monitoring result analysis reports to accurately determine whether the spillway structure is operating normally. Summary of the Invention

[0004] The purpose of this invention is to provide a vibration monitoring system for arch dam spillway, comprising a vibration sensor, a power supply module, a communication module, an arch dam spillway vibration measurement point layout design module, a vibration monitoring module, and a host computer. The vibration sensor is used to monitor the spillway vibration signal in real time and obtain the vibration monitoring signal. The power supply module is used to supply power to other modules. The communication module is used to transmit the vibration monitoring signal in real time. The arch dam spillway vibration measurement point layout design module is used to determine the installation location of the vibration sensor on site. The vibration monitoring module and the host computer are used to perform data calculation, analysis, and storage of the vibration monitoring signal.

[0005] Furthermore, the vibration sensor is a sensor based on fiber optic sensing technology, comprising a sensor unit, a signal processing and acquisition unit, a power supply unit, a central processing unit, a signal conversion unit, and a housing packaging unit; the sensor unit is connected to the signal processing and acquisition unit; the signal processing and acquisition unit is also connected to the central processing unit; the central processing unit is also connected to the signal conversion unit; the power supply unit supplies power to each unit in the vibration sensor; the sensor unit includes a low-frequency vibration displacement core; the signal processing and acquisition unit includes a preamplifier circuit, a filter circuit, and an analog-to-digital converter circuit; the housing packaging unit includes an internal base for each sensor module, a protective housing, and an adjustable mounting bracket.

[0006] Furthermore, the power supply unit includes a power supply interface, an AC-DC power supply circuit, and a DC-DC power supply circuit; the AC-DC power supply circuit includes a rectifier and a step-down regulator; the DC-DC power supply circuit includes a voltage conversion device and a voltage regulator; the power supply is connected to the AC-DC power supply circuit; the AC-DC power supply circuit is also connected to the signal processing unit and the DC-DC power supply circuit; the DC-DC power supply circuit is also connected to the signal acquisition unit, the central processing unit, and the signal conversion unit.

[0007] Furthermore, the central processing unit uses an STM32 controller as its control processor.

[0008] The purpose of this invention is to provide a method for identifying damage to a leakage structure, comprising: acquiring vibration monitoring signals and processing the vibration monitoring signals to obtain feature parameters; inputting the feature parameters into a support vector data description model, wherein the support vector data description model determines whether the vibration monitoring signals are normal; when the vibration monitoring signals are normal, the system issues a normal feedback; when the vibration monitoring signals are abnormal, the system issues an early warning.

[0009] Furthermore, the feature parameters include time-domain feature parameters and frequency-domain feature parameters; wherein, obtaining the feature parameters includes: performing frame-segmentation processing on the vibration monitoring signal to obtain multiple sub-vibration signals; processing the multiple sub-vibration signals respectively to obtain time-domain feature parameters and frequency-domain feature parameters; wherein, the time-domain feature parameters include root mean square, dual amplitude, variance, bias coefficient, and kurtosis coefficient; the frequency-domain feature parameters include centroid frequency and dominant frequency.

[0010] Furthermore, the support vector data description model is trained through a machine learning model, including: performing frame-by-frame processing on the collected historical vibration monitoring signals to obtain multiple historical sub-vibration signals; acquiring feature parameters of the multiple historical sub-vibration signals respectively; constructing target training samples based on the feature parameters, and obtaining labels for the target training samples based on the arch dam operating status corresponding to the historical vibration monitoring signals; inputting the target training samples into an initial support vector data description model, wherein the initial support vector data description model extracts normal feature domains of normal target training samples based on the labels; and using the initial support vector data description model with the normal feature domains as the vector data description model.

[0011] Furthermore, the constraint optimization function of the initial support vector data description model is:

[0012]

[0013] st ||x i -a|| 2 ≤R 2 +ξ i ξ i >0

[0014] Where F(R,a) represents the minimized error function; R represents the radius of the hypersphere; a represents the center of the hypersphere; C represents the penalty parameter that balances the hypersphere volume and the error rate; ξ i Let represent the slack variable; i represents the first variable, taking values ​​from 1 to n, where n is the total number of target training samples; st represents the constraint; x i This represents the i-th target training sample; ||*|| represents taking the absolute value.

[0015] Furthermore, the dual function of the constrained optimization function of the initial support vector data description model is:

[0016]

[0017]

[0018] Where F′ denotes the dual problem of minimizing the error function F(R,a); j represents the second variable, taking values ​​from 1 to n, where n is the total number of target training samples; α i Let x represent the i-th target training sample. i The corresponding Lagrange coefficient; α j Let x represent the j-th target training sample. j The corresponding Lagrange coefficients; x j K(x) represents the j-th target training sample; i ·x j) represents the kernel function of the initial support vector data description model.

[0019] Furthermore, it also includes: constructing a data report based on the judgment results output by the support vector data description model and their corresponding feature parameters; using the feature parameters in the data report as new target training samples and the judgment results corresponding to the feature parameters as the labels of the new target training samples to update the training sample library; and updating the support vector data description model based on the updated training sample library.

[0020] The technical solutions of the embodiments of the present invention have at least the following advantages and beneficial effects:

[0021] The vibration sensor used in the real-time monitoring system for the spillway vibration of ultra-high arch dams based on a novel vibration sensor is a new type of vibration sensor based on fiber optic sensing technology. It features easy installation, robust reliability, small size, high resolution, high measurement accuracy, strong anti-interference ability, strong stability, low power consumption, and long service life. It meets the requirements for long-term continuous operation in harsh on-site environments such as humid and dark environments, and solves the problem that traditional voltage-type sensors are easily damaged in humid spillway environments, leading to monitoring system failure. It also reduces the workload of subsequent system maintenance and lowers the system operation and maintenance costs.

[0022] The intelligent identification method for spillway structure damage based on monitoring system vibration data can learn from the spillway vibration data and continuously update and optimize the normal threshold of vibration indicators. As the amount of spillway vibration monitoring data increases, the vibration indicator threshold will continuously approach the true threshold, greatly improving the accuracy of forecasting and early warning of abnormal situations in the spillway structure. This solves the problem that traditional analysis methods often make incorrect judgments due to inaccurate threshold settings when judging the structural operating status by using fixed vibration data indicator thresholds, resulting in false alarms under normal conditions and no alarms under abnormal conditions.

[0023] The real-time vibration monitoring system for ultra-high arch dam spillway based on novel vibration sensors is highly integrated and intelligent. It can monitor the vibration characteristics of the spillway structure of ultra-high arch dams in real time. Combined with this intelligent damage identification method, the monitoring system can automatically generate vibration monitoring results analysis reports, accurately determine whether the spillway structure is operating normally, promptly detect operational defects and safety hazards in the spillway structure, take preventive measures, ensure the long-term safe and stable operation of the spillway structure of ultra-high arch dams, avoid engineering disasters, and has unparalleled advantages over other monitoring methods. Attached Figure Description

[0024] Figure 1 This is an exemplary schematic diagram of the arch dam spillway vibration monitoring system provided by the present invention;

[0025] Figure 2This is an exemplary schematic diagram of the vibration sensor provided by the present invention;

[0026] Figure 3 An exemplary flowchart of the method for identifying damage to a leakage structure provided by the present invention. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0028] Figure 1 This is an exemplary schematic diagram of the arch dam spillway vibration monitoring system provided by the present invention.

[0029] The arch dam spillway vibration monitoring system provided by this invention includes a vibration sensor, a power supply module, a communication module, an arch dam spillway vibration measurement point layout design module, a vibration monitoring module, and a host computer.

[0030] The vibration sensors are installed at corresponding locations on site according to the design of the vibration measurement points for the spillway of the ultra-high arch dam, and are used to monitor the spillway vibration signal in real time to obtain the vibration monitoring signal.

[0031] The power module is used to power other modules. It powers the vibration sensor and communication module. The sensor has a reserved power interface for external 220V AC voltage, with a maximum power supply current not exceeding 1.5A. The communication module mainly includes a mid-layer switch and a terminal server, powered by 220V AC.

[0032] The communication module is used for real-time transmission of vibration monitoring signals. It connects the vibration sensor, the vibration monitoring module, and the host computer to transmit vibration monitoring signals in real time. The sensor has a reserved SC fiber optic communication interface, which is connected to the fiber optic cable via a flange. Single-mode single-fiber transmission is used, and the monitoring data is transmitted to a fiber optic switch via fiber optic cable. The fiber optic switch aggregates all sensor measurement point interfaces in the field and connects to the terminal server via a network cable.

[0033] The arch dam spillway vibration monitoring point layout design module is used to determine the installation locations of the vibration sensors on site. These locations include horizontal displacement monitoring points for vibration sensors at the arch crown beam of the ultra-high arch dam, the sidewalls of the dam gallery, and near the spillway openings; vertical displacement monitoring points for vibration sensors at approximately 15-meter intervals in the water tongue area of ​​the energy dissipation facility downstream of the dam; and one to two sensors at the end of the dam foundation curtain grouting tunnel or drainage tunnel as free-field monitoring points for comparative analysis.

[0034] The vibration monitoring module and the host computer are used to perform data calculation, analysis and storage on the vibration monitoring signals.

[0035] For example, such as Figure 1 As shown, the arch dam spillway vibration monitoring system includes vibration sensors 10, an arch dam spillway vibration measurement point layout design module 20, a power supply and communication module 30, a vibration monitoring module, and a host computer 40. Vibration sensors 10 are arranged and installed at corresponding locations on the spillway structure of the ultra-high arch dam according to the arch dam spillway vibration measurement point layout design module 20 to monitor spillway vibration signals. The power supply and communication module 30 provides the sensors with 220V AC power and network communication, uploading the signals collected by the sensors to the vibration monitoring module and the host computer 40 for calculation, storage, intelligent damage identification and analysis, display, report generation, and alarm functions. The vibration monitoring module contains vibration monitoring system software.

[0036] The arch dam spillway vibration monitoring point layout design module 20 includes a dam body sensor layout design 201 and an energy dissipation area sensor layout design 202. In this embodiment, the dam body sensor layout design 201 involves arranging new vibration sensors at the arch crown beam of the ultra-high arch dam, the sidewalls of the dam gallery, and near the spillway openings. The energy dissipation area sensor layout design 202 involves arranging new vibration sensors at approximately 15-meter intervals in the water tongue area of ​​the energy dissipation facility downstream of the dam. Additionally, one to two sensors are arranged at the ends of the dam foundation curtain grouting tunnel or drainage tunnel as free-field monitoring points for comparative analysis.

[0037] The power and communication module 30 includes an optical fiber 301, a switch 302, a communication interface 303, a server 304, and a power interface 305. The optical fiber 301 is a single-mode optical fiber. The switch 302 is a gigabit fiber optic switch with at least two gigabit RJ45 ports and twelve gigabit SFP optical ports, each equipped with an optical module. The communication interface 303 mainly provides communication interfaces between the optical fiber and the new vibration sensor, the switch, and the server. The optical fiber communication interface is connected to the new vibration sensor communication interface 110 via a flange. The switch 302 is connected to the server 304 via a network cable. The server 304 contains the system software and a host computer. The power interface 305 provides 220V AC power to the vibration sensor 10, switch 302, server 304, and other devices.

[0038] The vibration monitoring module and host computer 40 use LabVIEW as the development and operation platform. It includes a main control module 401, a data display module 402, a data storage module 403, a data analysis module 404, a report generation module 405, and an alarm module 406. The main control module 401 controls and operates the entire monitoring system, controlling the system to start or stop collecting discharge vibration data according to the actual flood discharge conditions and the operational needs of the ultra-high arch dam. The data display module 402 is used for human-computer interaction, displaying vibration monitoring data and waveforms. The data storage module 403 stores the discharge vibration monitoring data, with storage paths including a cloud database and a server hard drive. The data analysis module 404 performs calculations and analyses using a structural damage intelligent identification method based on the monitoring system's vibration data. The report generation module 405 generates a discharge safety evaluation report for the ultra-high arch dam's discharge structure based on a pre-set report template and the calculation and analysis results from the data analysis module 404. The alarm module 406 determines whether to alarm based on the normal characteristic range of the vibration signal of the ultra-high arch dam spillway structure, which is continuously and automatically updated and improved by the data analysis module 404. If the vibration signal exceeds the normal characteristic range, a red alarm label is output.

[0039] Figure 2 This is an exemplary schematic diagram of the vibration sensor provided by the present invention.

[0040] The vibration sensor provided by this invention is a fiber optic sensing technology-based sensor used to collect vibration signals of the spillway structure during flood discharge at ultra-high arch dams in real time. The vibration sensor includes a sensor unit, a signal processing and acquisition unit, a power supply unit, a central processing unit, a signal conversion unit, and a housing packaging unit; the sensor unit is connected to the signal processing and acquisition unit; the signal processing and acquisition unit is also connected to the central processing unit; and the central processing unit is also connected to the signal conversion unit.

[0041] The power supply unit supplies power to each unit in the vibration sensor. The power supply unit includes a power supply interface, an AC-DC power supply circuit, and a DC-DC power supply circuit; the AC-DC power supply circuit includes rectifiers and step-down regulators; the DC-DC power supply circuit includes voltage conversion devices and voltage regulators; the power supply is connected to the AC-DC power supply circuit; the AC-DC power supply circuit is also connected to the signal processing unit and the DC-DC power supply circuit; the DC-DC power supply circuit is also connected to the signal acquisition unit, the central processing unit, and the signal conversion unit.

[0042] The sensor unit includes a low-frequency vibration displacement core.

[0043] The signal processing and acquisition unit includes a preamplifier circuit, a filter circuit, and an analog-to-digital converter circuit. The analog-to-digital converter circuit is implemented using a bipolar high-speed, high-resolution ADC chip.

[0044] The housing encapsulation unit includes an internal base for each sensor module, a protective housing, and an adjustable mounting bracket. The protective housing is made of stainless steel plate and is printed with the sensor's IP address, serial number, and type identifier (vertical or horizontal displacement sensor).

[0045] The central processing unit uses an STM32 controller as its control processor.

[0046] For example, such as Figure 2As shown, the vibration sensor 10 includes a power interface 101, an AC-DC power supply unit 102, a DC-DC power supply unit 103, a power isolation unit 104, a vibration core 105, a signal processing unit 106, a data acquisition unit 107, a central processing unit 108, a signal conversion unit 109, and a communication interface 110. The power interface 101 is connected to an external 220V AC power supply to power the sensor. The AC-DC power supply unit 102 includes a rectifier circuit and a step-down regulator circuit, which rectifies and steps down the input 220V AC power supply voltage to output ±12V DC, powering the signal processing unit 106. Simultaneously, a 12V DC output is connected to the DC-DC power supply unit 103. The DC-DC power supply unit 103 includes a voltage conversion circuit and a voltage regulator circuit, which steps down and regulates the 12V DC, outputting 5V DC to the power isolation unit 104 for conversion into a fully isolated 5V DC, powering the data acquisition unit 107 and the central processing unit 108. The vibration signal generated by the vibrator 105 due to the on-site leakage condition is transmitted to the signal processing unit 106. After amplification and filtering by the signal processing unit 106, it is transmitted to the data acquisition unit 107. The data acquisition unit 107 acquires and converts the amplified analog vibration signal from analog to digital, and outputs a digital signal to the central processing unit 108 for processing, temporary storage, conversion, and output. In this embodiment, the vibration amplitude monitoring sensitivity of the vibrator 105 is 5mV / µm, the range is ±2.5mm, and the vibration frequency response range is 0~200Hz. The signal processing unit 106 includes a signal amplification circuit and a filtering circuit. The signal amplification circuit is implemented using a low-power operational amplifier circuit with an amplification factor of up to 15 times, and the filtering circuit is implemented using an RC filter method. The data acquisition unit 107 mainly consists of an A / D conversion circuit. In this embodiment, the A / D conversion circuit uses the AD7606 digital-to-analog converter chip from Analog Devices (ADI), which has a 16-bit high resolution, supports bipolar signal input ranges of ±5V or ±10V, and each channel can maintain a high sampling rate of 200kSPS. The central processing unit 108 mainly consists of a core processor and its peripheral circuits. In this embodiment, the core processor used is the STM32F103C8T6 manufactured by ARM, and its high-speed GPIO peripheral interface is connected to the data acquisition unit 107. The signal conversion unit 109 mainly consists of a photoelectric conversion circuit, which converts vibration data from electrical signals to optical signals for transmission. It is connected to an external optical fiber through the communication interface 110, and after being aggregated by a mid-level switch, it is transmitted to the vibration monitoring module and the host computer 40. The vibration monitoring module is equipped with a vibration monitoring system.

[0047] Figure 3 An exemplary flowchart of the method for identifying damage to a leakage structure provided by the present invention.

[0048] like Figure 3 As shown, the method for identifying damage to a leakage structure provided by the present invention includes:

[0049] Vibration monitoring signals are acquired and processed to obtain feature parameters. Support Vector Data Description (SVDD) is then used to describe the features of the vibration monitoring signals acquired by the monitoring system.

[0050] The feature parameters include time-domain feature parameters and frequency-domain feature parameters; wherein, obtaining the feature parameters includes: performing frame-segmentation processing on the vibration monitoring signal to obtain multiple sub-vibration signals; processing the multiple sub-vibration signals respectively to obtain time-domain feature parameters and frequency-domain feature parameters; wherein, the time-domain feature parameters include root mean square, dual amplitude, variance, deviation coefficient, and kurtosis coefficient; the frequency-domain feature parameters include centroid frequency and dominant frequency.

[0051] The feature parameters are input into the support vector data description model, which then determines whether the vibration monitoring signal is normal.

[0052] The support vector data description model is trained using a machine learning model and includes: framing the collected historical vibration monitoring signals to obtain multiple historical sub-vibration signals; acquiring feature parameters for each of the multiple historical sub-vibration signals; constructing target training samples based on the feature parameters, and obtaining labels for the target training samples based on the arch dam operating status corresponding to the historical vibration monitoring signals. For example, labels for normal or abnormal states are manually assigned based on the site conditions; the target training samples are input into an initial support vector data description model, which extracts the normal feature domain of normal target training samples based on the labels; and the initial support vector data description model with the normal feature domain is used as the vector data description model. By collecting vibration data, extracting features, and determining normal / abnormal states under different operating conditions during long-term discharge operation of ultra-high arch dams, the normal feature domain of vibration data is continuously updated and ultimately formed. An alarm is triggered when the vibration data features exceed the existing feature domain.

[0053] It also includes: constructing a data report based on the judgment results output by the support vector data description model and their corresponding feature parameters; using the feature parameters in the data report as new target training samples and the judgment results corresponding to the feature parameters as the labels of the new target training samples to update the training sample library; and updating the support vector data description model based on the updated training sample library.

[0054] The constraint optimization function of the initial support vector data description model is:

[0055]

[0056] st ||xi -a|| 2 ≤R 2 +ξ i ξ i >0

[0057] Where F(R,a) represents the minimized error function; R represents the radius of the hypersphere; a represents the center of the hypersphere; C represents the penalty parameter that balances the hypersphere volume and the error rate; ξ i Let represent the slack variable; i represents the first variable, taking values ​​from 1 to n, where n is the total number of target training samples; st represents the constraint; x i This represents the i-th target training sample; ||*|| represents taking the absolute value.

[0058] The dual function of the constrained optimization function of the initial support vector data description model is:

[0059]

[0060]

[0061] Where F′ denotes the dual problem of minimizing the error function F(R,a); j represents the second variable, taking values ​​from 1 to n, where n is the total number of target training samples; α i Let x represent the i-th target training sample. i The corresponding Lagrange coefficient; α j Let x represent the j-th target training sample. j The corresponding Lagrange coefficients; x j K(x) represents the j-th target training sample; i ·x j ) represents the kernel function of the initial support vector data description model. By solving the constraint optimization function using the dual function of the constraint optimization function, the normal feature domain of normal data can be obtained through the kernel function K(*), thus transforming the nonlinear problem in low-dimensional space into a linear problem in high-dimensional space. This greatly enhances the robustness and generalization performance of the SVDD (Support Vector Data Description) model and enriches the application scenarios of the SVDD model.

[0062] When the vibration monitoring signal is normal, the system issues a normal feedback; when the vibration monitoring signal is abnormal, the system issues an early warning.

[0063] For example, the normal and abnormal vibration signals of an ultra-high arch dam under specific working conditions are simulated using two measured structural vibration signals, and the detailed implementation process of SVDD in the intelligent structural damage identification method based on vibration data of the monitoring system is explained.

[0064] First, the collected vibration signals are segmented into frames of appropriate length to form multiple vibration signal sample data. Then, multiple characteristic parameters in the time and frequency domains of each signal sample are calculated. Time-domain characteristic parameters include root mean square (RMS), amplitude, variance, skewness coefficient, and kurtosis coefficient, reflecting the waveform similarity of the same signal or the correlation between different signals in the vibration of the spillway structure, such as the autocorrelation function and cross-correlation function. Frequency-domain characteristics include centroid frequency and dominant frequency. The selected characteristic parameters should reflect the operating state of the ultra-high arch dam under different working conditions during long-term spillway operation. For example, the RMS index can reflect the intensity of vibration under the pulsating pressure of water flow, characterizing the strength of the vibration; the dominant frequency can reflect the random vibration characteristics of the structure in the frequency domain; and the kurtosis coefficient and skewness coefficient are used to determine the normality of the vibration sequence of the spillway structure. Assuming the number of characteristic parameters is d, a set of target training samples with d-dimensional features can be obtained for each vibration signal sample data. After the target training samples are labeled, they are used to train the SVDD model. When the training samples contain both normal and abnormal training samples, the normal feature domain of the vibration signal of the spillway structure of the ultra-high arch dam can be obtained. This feature domain can be gradually improved as the number and types of target training samples become richer, and finally form a feature domain that can completely characterize the normal and abnormal vibration conditions of the spillway structure of the ultra-high arch dam.

[0065] In this simulation, training samples representing normal conditions were used as positive samples, and samples representing abnormal conditions were used as negative samples. All samples were divided into training and test sets in a 4:1 ratio. The trained SVDD model achieved a classification accuracy of 98.57% on the training set and 98.18% on the test set. Principal component analysis (PCA) was used to reduce the d-dimensional features of the training samples to 2-dimensionality, enabling visualization of the hypersphere distance cloud map and classification decision surface.

[0066] A specific implementation was applied to a 300m-class ultra-high arch dam. One sensor was deployed near each of the four spillway outlets on the dam crest. Eighteen sensors were deployed inside the dam body at EL.1829m, EL.1778m, EL.1730m, and EL.1664m along the gallery walls and the arch crown beam. Fifty sensors were deployed in the downstream stilling basin floor gallery area at approximately 15-meter intervals (horizontal and vertical). One sensor was also deployed as a free-field measuring point on each of the left and right bank curtain tunnel walls at EL.1829m. The sensors were fixed using adjustable brackets, which were secured to designated positions on the sidewalls with expansion bolts. The tilt angle of the sensors during installation did not exceed 1.5°. Power supply equipment and switches were installed in the dam body and stilling basin observation rooms, with power cables connecting to the sensors and single-mode optical fibers for communication. After the sensor collects the leakage vibration signal, it transmits it to the switch through optical fiber. After the switch completes the signal aggregation, it transmits it to the server through network cable. The monitoring system software and host computer perform data calculation, storage, analysis, display, alarm, etc.

[0067] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for identifying damage to a bleed-out structure, characterized in that, This invention relates to a vibration monitoring system for arch dam spillways. The system comprises vibration sensors, a power supply module, a communication module, a vibration measurement point layout design module, a vibration monitoring module, and a host computer. The vibration sensors are used to monitor spillway vibration signals in real time, generating vibration monitoring signals. The vibration sensors are based on fiber optic sensing technology and include sensor units, signal processing and acquisition units, a power supply unit, a central processing unit, a signal conversion unit, and a housing packaging unit. The sensor units are connected to the signal processing and acquisition units, which are also connected to the central processing unit. The central processing unit is also connected to the signal conversion unit. The power supply unit powers each unit within the vibration sensor. Each sensor unit includes a low-frequency vibration displacement core. The signal processing and acquisition unit includes a preamplifier circuit, a filter circuit, and an analog-to-digital converter circuit. The housing packaging unit includes an internal base for each sensor module, a protective housing, and an adjustable mounting bracket. The power supply module powers other modules. The communication module transmits vibration monitoring signals in real time. The arch dam spillway vibration measurement point layout design module determines the installation location of the vibration sensors on-site. The vibration monitoring module and the host computer are used to perform data calculation, analysis and storage on the vibration monitoring signals; The method for identifying damage to the leakage structure includes: Vibration monitoring signals are collected and processed to obtain characteristic parameters; The feature parameters are input into a support vector data description model (SVM), which determines whether the vibration monitoring signal is normal. The SVM is trained using a machine learning model and includes: framing the collected historical vibration monitoring signal to obtain multiple historical sub-vibration signals; acquiring feature parameters for each of the multiple historical sub-vibration signals; constructing a target training sample based on the feature parameters and obtaining a label for the target training sample based on the arch dam operating state corresponding to the historical vibration monitoring signal; inputting the target training sample into an initial SVM, which extracts the normal feature domain of the normal target training sample based on the label; using the initial SVM with the normal feature domain as the SVM; constructing a data report based on the judgment result output by the SVM and its corresponding feature parameters; updating the training sample library by using the feature parameters in the data report as new target training samples and the judgment result corresponding to the feature parameters as the label of the new target training samples; updating the SVM based on the updated training sample library; issuing a normal feedback when the vibration monitoring signal is normal; and issuing an early warning when the vibration monitoring signal is abnormal.

2. The method for identifying damage to a leakage structure according to claim 1, characterized in that, The feature parameters include time-domain feature parameters and frequency-domain feature parameters; wherein, obtaining the feature parameters includes: The vibration monitoring signal is processed by framing to obtain multiple sub-vibration signals; The multiple sub-vibration signals are processed to obtain time-domain characteristic parameters and frequency-domain characteristic parameters; wherein, the time-domain characteristic parameters include root mean square, double amplitude, variance, deviation coefficient and kurtosis coefficient; the frequency-domain characteristic parameters include centroid frequency and dominant frequency.

3. The method for identifying damage to a bleed-out structure according to claim 1, characterized in that, The constraint optimization function of the initial support vector data description model is: in, This represents minimizing the error function; R represents the radius of the hypersphere; denoted by ; C represents the center of the hypersphere; C represents the penalty parameter that balances the hypersphere's volume and the error rate. represents the slack variable; i represents the first variable, taking values ​​from 1 to n, where n is the total number of target training samples; Indicates constraint; This represents the i-th target training sample; This indicates taking the absolute value.

4. The method for identifying damage to a bleed-out structure according to claim 3, characterized in that, The dual function of the constrained optimization function of the initial support vector data description model is: in, Represents minimizing the error function The dual problem; j represents the second variable, taking values ​​of 1, ..., n, where n is the total number of target training samples; Represents the i-th target training sample The corresponding Lagrange coefficients; Represents the j-th target training sample The corresponding Lagrange coefficients; This represents the j-th target training sample; This represents the kernel function that describes the model based on the initial support vector data.

5. The method for identifying damage to a bleed-out structure according to claim 1, characterized in that, The power supply unit includes a power supply interface, an AC-DC power supply circuit, and a DC-DC power supply circuit. The AC-DC power supply circuit includes rectifiers and step-down regulators; The DC-DC power supply circuit includes voltage conversion devices and voltage regulators; The power supply is connected to the AC-DC power supply circuit; the AC-DC power supply circuit is also connected to the signal processing unit and the DC-DC power supply circuit; the DC-DC power supply circuit is also connected to the signal acquisition unit, the central processing unit, and the signal conversion unit.

6. The method for identifying damage to a bleed-out structure according to claim 1, characterized in that, The central processing unit uses an STM32 controller as its control processor.

Images